Uproot Subclasses

[BenGalewsky]

Problem

Fixes #20
For testing and development it's much more convenient to run an analysis locally using uproot instead of in spark. We need a way to run the exact same code on spark as locally with just a config change.

#Approach
Pulled the spark-specific stuff into a new class called an Executor which is set at config time. Created a new dataset based on uproot reader. Made the select_columns method gather up the dictionary of nested arrays and generated a native JaggedArray dictionary. This becomes a new dataset subclass, AwkwardDataset

The UDFs operate on the AwkwardDataset.

The FileDatasetManager depended on an instance of spark for querying datasets. Made a new subclass which just keeps the dataset definitions in a python dictionary.